PIPE AUTOMATED INSPECTION HELP

[coolrthancoolio]

Hi everyone

My Team and I are developing a vision system that performs a pipe inspection for our senior project. At this time, we are only using a 1.3 megapixel usb webcam for processing but will test more available cameras. I will attach raw images that we have inspected. So far, we are in the developing stages. We are using the LabView Vision Builder for Automated Inspection. I have attached the .vbai folder to this post. We are having a difficulty indicating if our overall image has either passed or failed when it is ran once or continous. It always indicates that the image has passed.

Questions:

1. If we obtain a higher quailty camera would it require less filtering of an image?

2. Can anyone give us tips or comments on the steps we used to process the image?

3. Is there a way to save all failed images for further inspection?

4. Is there anyone that has created something similar?

5. Are there any tutorials that will help us out?

Any input is greatly appreciated.  Thanks.

[wbadry]

First of all, I'd love to wish you all the best in your project 

Concerning your questions, I may answer it up to the best of my knwoledge

1. If we obtain a higher quailty camera would it require less filtering of an image?

It depends on the choice of your camera (Vendor and Resolution). Reputable vendors offer cameras for research and industry that has low noise. Meanwhile, the higher resolution you get, the more features and details in the image you can detect.

2. Can anyone give us tips or comments on the steps we used to process the image?

Would you state in details what do you want are your desired objective (inspection on color, cracks, corrosion)

3. Is there a way to save all failed images for further inspection?

I know how to do it in LabVIEW and VB.NET. Unfortunately not aware of VBAI

4. Is there anyone that has created something similar?

I carried out inspection for cracks and quality of shapes on garnished ceramic plates.

5. Are there any tutorials that will help us out?

I think reading the help is one of the best tutorials you would ever find.

VBAI Tutorials

VBAI October 2010

Customizing the Inspection Interface in NI Vision Builder for Automated Inspection

Also see this thread

Hope it helps 

[Brad]

It's not clear to me what you want to find out from those images. Are you just looking for the holes in the plates/pipes? Maybe you could describe in more detail what you are looking for and what is considered passing/failing. I would recommend looking at the detect objects step where you can threshold the color image, select particle filtering options (based on size, location, shape, etc.) and pass/fail based on the number of objects foundd. I included a modified inspection to illustrate how to do this with the image logging step (I made it so passed images get logged to one folder and failed images get logged to another). You can modify this step to only log failed images if you prefer. I aldo didn't add any of the more complex filtering criteria available in the Detect Object step (wasn't needed to find the holes), but you can play with the filtering criteria if you need more granularity in how particles get filtered out. I would say you only need more resolution if you can't solve the problem with your current resolution.

Hope this helps,

Brad

[Diljay]

First of all, I would like to say thank you for your input and speedy reply.

Our main goals for this project are to:

1. locate cracks and determine whether or not to fail or pass based on the length and crack separation.

2. locate surface defects (corrosion and dents).  Assuming our monochrome camera is in a fixed posistion with a consistent light source, how do we differentiate between corrosion and dust if they share the same intensities?

3. Establish a coordinate system to determine the location of the crack on the pipe.  How do we establsih a coordinate system if the camera is moving vertically up the pipe while revolving around the diameter.

---

@VBCoder wrote:

4. Is there anyone that has created something similar?

I carried out inspection for cracks and quality of shapes on garnished ceramic plates.

---

I know your algorithm is way more complex than ours but can you give us some input please

So far we've acquired an image from a webcam (temporary for now).  The image is just a white paper with lines drawn on to simulate cracks.

The process goes as follows

1. Color plane extraction to convert to greyscale

2. Threshold to convert to binary

3. Removing small particles with 2 or three iterations

By doing this, the program shows distinct lines but we do not know the neccessary steps to filter the image out further. 

I know we have to input some form of measurement to we can set value limits for the size of the crack.

I am almost finished with the tutorials and will try alternative methods.

Thanks

[Diljay]

Thank you for the VI brad.  It will come in handy.